Creating scatter plots using geom_point()

We also created our first ggplot() object to explore the question

1. Is there a relationship between healthcare coverage and healthcare spending in the United States?

hc %>%
  filter(type == "Employer", year == "2013") %>% 
  ggplot(aes(x = spending_capita, y = prop_coverage)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("coverage proportion") + 
  geom_smooth(method = "lm", col = "red")

It would be nice to know which state is represented by which state. For this, we will introduce another geom called geom_text().

Adding layers of text using geom_text()

In our dataset, we have information about the abbreviation for each state. We could add the abbreviations for each state next to the point on the plot to assess which states have a higher or lower coverage for a given amount of money they spend per capita.

hc %>% 
  filter(type == "Employer", year == "2013") %>% 
  ggplot(aes(x = spending_capita, y = prop_coverage)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("coverage proportion") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb))

That is cool, but it would be even better if we could nudge the text over a bit. Let’s look at the help file for geom_text():

?ggplot2::geom_text

We see there is an argument called nudge_x and nudge_y. We can use these to nudge the text over a bit so the text is not directly on top of the points.

hc %>% 
  filter(type == "Employer", year == "2013") %>% 
  ggplot(aes(x = spending_capita, y = prop_coverage)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("coverage proportion") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb), nudge_x = 150)

## add your code here

faceting across a variable

Ok, getting back to our original question:

1. Is there a relationship between healthcare coverage and healthcare spending in the United States?

We saw there was a positive relationship, but this was only for one type of healthcare coverage (Employer) and one year. What about the other types?

For this, we will introduce facets. The idea of faceting is to stratify the data by some variable and make the same plot for each strata.

For example, if we wanted to facet by the type variable, we will add a layer to our ggplot() object using the facet_grid() or facet_wrap() functions. The function expects the row and column variables to be separated by a ~.

hc %>%
  filter(year == "2013") %>% 
  ggplot(aes(x = spending_capita, y = prop_coverage, 
             color = region)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("coverage proportion") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb), nudge_x = 150) + 
  facet_wrap(~type)

## Warning: Removed 9 rows containing non-finite values (stat_smooth).

## Warning: Removed 9 rows containing missing values (geom_point).

## Warning: Removed 9 rows containing missing values (geom_text).

We see that the proportion of people covered have different scales in the y-axis. Let’s read the help file to see if there is some way to not restrict the y-axis to be the same.

?ggplot2::facet_grid

Yes, we see there is an argument called scales that can be free_y, (free columns), free_x (free rows), and free (both). Let’s try free_y and look at a different year (year=="2014"):

hc %>%
  filter(year == "2014") %>% 
  ggplot(aes(x = spending_capita, y = prop_coverage, 
             color = region)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("coverage proportion") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb), nudge_x = 150) + 
  facet_wrap(~type, scales="free_y")

## Warning: Removed 10 rows containing non-finite values (stat_smooth).

## Warning: Removed 10 rows containing missing values (geom_point).

## Warning: Removed 10 rows containing missing values (geom_text).

Given we know Other Public refers to the military or Veterans Adminstration, we can see states like HI, VA, NV have a larger proportion of military or VA Other Public type coverage. While a state like AK has a similar proportion of Other Public coverage, it has a much larger spending per capita.

We also see a negative relationship with the Uninsured type. The more states spend, the less uninsured people in the state.

Creating boxplots using geom_boxplot()

Next, let’s revisit the second question.

2. Which US states spend the most and which spend the least on healthcare? How does the spending distribution change across geographic regions in the United States?

Let’s try making a boxplot with ggplot2. If you recall, the way to do this in base R was:

boxplot(hc$spending_capita ~ hc$region)

Now, we introduce the geom_boxplot() function. Note, we needed to tell ggplot2 what needs to be along the x and y axis in aes().

hc %>% 
  ggplot(aes(x = region, y = spending_capita)) + 
  geom_boxplot()

faceting by two variables

3. How do healthcare coverage and spending relate to life expectancy?

In our last section, let’s first explore the relationship between spending and life expectancy using facets. We were faceting by healthcare type, so let’s stick with that, but change the y-axis to life_exp_years.

p <- hc %>%
  filter(year == "2014") %>% 
  ggplot(aes(x = spending_capita, y = life_exp_years, 
             color = region)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("life expectactancy") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb))

p + facet_wrap(~type, scales="free")

Hmm… So looks like there is a positive association for all the types (more you spend the longer the life expectacy). But there is something strange about the fact that the regions have very different life expectancies.

Let’s try faceting by both region and type. Note that we can facet by rows putting a column name before the ~ and facet by columns putting a column name after the ~. We are also using facet_grid() instead of facet_wrap().

p + facet_grid(region~type, scales="free")

What trends do you see?

Finally, let’s do the same for the prop_coverage column instead of spending_capita column to explore the relationship between life expectancy

p <- hc %>%
  filter(year == "2014") %>% 
  ggplot(aes(x = prop_coverage, y = life_exp_years, 
             color = region)) + 
  geom_point() + 
  xlab("spending per capita") + 
  ylab("life expectactancy") + 
  geom_smooth(method = "lm", col = "red") + 
  geom_text(aes(label=abb))

p + facet_grid(region~type, scales="free")

## Warning: Removed 10 rows containing non-finite values (stat_smooth).

## Warning: Removed 10 rows containing missing values (geom_point).

## Warning: Removed 10 rows containing missing values (geom_text).

What trends do you see? How would you summarize these results? What’s most interesting?

ggplot2 cheatsheet